tic web technologies for big data analysis and the
need for standardized test beds as baselines. For
instance, establishing semantic interoperability
via alignments between different standards and
data formats in the oil and gas industry was discussed as an interesting test bed for the suitability
of semantic technologies.
The fourth topic was concerned with the combination of deduction and induction approaches.
There was a broad agreement that an efficient
handling of big data has to enrich bottom-up, data-driven methods known from data mining and
machine learning with a declarative top-down
layer provided by various semantic web technologies as well as ontologies that provide formal specifications of constraints and background knowledge. IBM’s Watson was discussed as one example
where knowledge and large amounts of unstructured data came together. Along the same lines,
the participants discussed current trends in learning axioms from data as well as the increasing interest of social network researchers in semantic
Finally, the fifth topic dealt with big data integration. The participants noted that the focus on
storing, accessing, and processing big data has
drawn attention away from equally important
challenges such as how to search for, normalize,
transform, aggregate, integrate, and summarize
massive amounts of data from highly heterogeneous sources. This, however, is a prerequisite to
arrive at a more holistic understanding of many
major questions in research and society that typically span across multiple domains — global
change being a typical example that involves data and theories from economics, political sciences,
geography, climatology, and so forth.
While most of the symposium focused on semantic technologies, one interactive session was
entirely devoted to the role of ontologies. More
concretely, the participants discussed the sweet
spot for ontologies as a communication layer between numerical and statistical methods and
models on the one hand, and the human user on
the other hand. This included questions about the
relation of semantic web ontologies to the ANSI
data model stack as well as the appropriate level
and granularity of axiomatization. Following the
well known slogan that a little semantics goes a
long way, the participants debated the question of
how much a little actually is.
The symposium also included presentations by
three invited speakers. Jennifer Golbeck from the
University of Maryland presented her work on
how to compute trust and building trust with
users’ social media data. Peter Fox from the Rens-
selaer Polytechnic Institute reported on the medi-
ation and integration of geodata through formal
vocabularies and highlighted the role of semantic
heterogeneity. Finally, Stanford’s Michel Dumontier dis-
cussed the generation of biomedical hypotheses using se-
mantic web technologies. This last keynote was organ-
ized jointly with the Discovery Informatics symposium
and was followed by a joint discussion session.
Frank van Harmelen, James A. Hendler, Pascal Hitzler,
and Krzysztof Janowicz served as cochairs of this sympo-
Here’s to all those who
think outside the box.
In our case,
the search box.
We’re integrating Bing into
the products you use every
day so you can search your
world, not just the web.